The MOA interface

We download MOA and run it. Incremental data stream mining calls for different evaluation methods from batch operation. One possibility is to interleave training and testing by periodically holding out some test data from the input stream (called “periodic” evaluation); another is to test the current classifier on each new instance in the stream before using it for training (called “prequential” evaluation). MOA, which can be invoked through its interactive interface or from the command line, includes many data stream generators. Here we run the HoeffdingTree algorithm on data from the HyperplaneGenerator, and evaluate it both periodically and prequentially.